Of great interest to forecasters of the economy is predicting the “business cycle”, or the overall level of economic activity. The business cycle affects society as a whole by its fluctuations in economic quantities such as the unemployment rate (the misery index), corporate profits (which affect stock market prices), the demand for manufactured goods and new housing units, bankruptcy rates, investment in research and development, investment in capital equipment, savings rates, and so on. The business cycle also affects important socio-political factors such as the the general mood of the people and the outcomes of elections. A scientific model of business cycle dynamics is not yet available due to the complexities of the economic system, the impossibility of doing controlled experiments on the economy, and the nonquantifiable factors such as mass psychology and sociology which influence economic activity. Given the absence of reliable or convincing scientific models of the business cycle, economists have resorted to analyzing and forecasting economic activity by using the empirical “black box” techniques of standard linear time series analysis. We have developed robust predictive models of the business cycle based on neural networks which outperform the standard linear AR models used by most economists. Economic statistics for the U.S. such as the national income and product accounts and the indices of leading, coincident, and lagging indicators have been collected and computed by the Bureau of Economic Analysis of the Department of Commerce since 1946. The standard measures of economic activity used by economists to track the business cycle are the Gross Domestic Product (GDP)1 and the Index of Industrial Production (IP). GDP is a broader measure of economic activity than is IP. However, GDP is computed by the Department of Commerce on only a quarterly basis, while Industrial Production is computed and published monthly. We have focussed on the Index of Industrial Production rather than GDP for three reasons. First, being published monthly, there is more data available for Industrial Production than for GDP. Second, the IP series is more timely than GDP and is therefore watched more closely by business, financial, and The authors are with Nonlinear Prediction Systems, PO Box 681, Portland, OR 97207-0681 and the Department of Computer Science of the Oregon Graduate Institute, PO Box 91000, Portland, OR 97291-1000. Email correspondence may be directed to moody@cse.ogi.edu. 1In 1990, GDP replaced Gross National Product (GNP) as a standard measure of domestic economic activity. GNP includes so-called “factor payments” to and “factor income” from foreign sources which are not included in GDP. These factors relate to interest, dividends, and reinvested earnings by foreign subsidiaries of US companies. As such, they are not really part of the domestic economy. GDP also includes the consumption of fixed capital, and important effect which is not captured by GNP.